The difficulty in controlling the proper bonding during manufacturing sandwich for ship structure is a major problem. One of the damage types that often occurs is debonding caused by the lack of the adhesive layer bond between the two faceplate materials and the sandwich core at the centre. Furthermore, the identification of damage in ship structure becomes essential to ensure the structural integrity and security issues. Based on the issues above, damage identification on the sandwich plate of Ferry Ro-Ro with several debonding ratios using the k-NN algorithm was proposed. The cross-validation method was used to check the validity of the proposed algorithm model in practice. The proposed models were analyzed with various k-fold values, k-NN values, and number of iterations to investigate the influence of the those parameters on the accuracy of the model. From the result, it can be found that increasing k-fold value is the best strategy to increase the accuracy of the model. However, increase number of neighbors (k) caused a bias so that the accuracy decreased. The tests resulted in the highest accuracy of 97.60% obtained at k-fold 11 with k = 3 and 90 iterations. It can be summarized that the large number of k-folds and small k-NN parameters do not require large iterations to obtain high prediction results.